Running immich since 2 years, never had a single breaking issue or instability of any kind.

There is really no reason not to choose immich unless you prefer something that’s only in ente.

Thank you, i rhink i will the subwave handle the TTS with kokoro, i just need to understan how to add an italian spekaer and i am good to go.

Its fun, it’s original, it’s a learning experience. what else would one need?

A few personal notes here https://wik/ i.gardiol.org/3-linux-gentoo see the su pages as well

Gentoo is a great choice! I use it an all my devices from servers to laptops and workstations . It’s fun, solid, stable and efficient.

Good choice, and… Good luck… Dm for any issue.

The server is plugged to a network Kwm so there is an actual output. And it was working just fine even without anything plugged in, I can confirm. But the nkwm is just practical

Yes,but that is a headless server anyway. So I will indeed switch to using the igpu as primary video sooner or later (need to physically go to the server, I have a network Kem but that would be pretty useless after I get into bios and change the video output)

Yes. All you say make sense. Tts can and shall go on CPU, who cares.

It’s handling both general chat and ComfyUI the biggest point, but to be honest o don’t need them both at the same time. The point is how easy is to switch since this is an unattended server, shutting down containers might not be the easiest approach. Specially if some other family member want to use it.

I will switch screen out on the igpu, it’s unused anyway, and see if I can run one card only, but I am struggling in running almost any model on LocalAI with llama.cpp I need to dig in more time on the issue.

Adding the second Nvidia will require a new psu, the mobo has the slot, even if it’s only 16x downgraded to 4x according to Claude it will work just fine for normal inference. But I will keep it as a last resort, I don’t particularly enjoy fiddling with that bios.

In any case I want both the llm and comfy to run on GPU and not offload so CPU.

I gave it a spin today and pretty happy so far. Took a bit to run, llama.cpp got installed broken and had to reinstall it …

But it pretty much replaced my entire openwebui+ollama stack with something seemingly much more capable and open.

Not clear to me where to find Qwen 3.6 35B-A3B, at Q4_K_M, and if that will leave free vram for smaller models (tts)

I am running LocalAI that manages for me llama.cpp plus at least another backend for the tts and I cannot find that model on the list, but it can import models downloaded manually

I am trying to understand if it make sense to run two cards.

I am using ai for chat and to support my subwave radio station (ai dj + music tagging and text to speech, small models are fine) so I am going to test a qwen3.5 9b derived model to have free ram to run text2speech model as well.

The second usage for ai is image generation, I am using ComfyUI so it’s not clear to me how that plays with the other models… Will they be unloaded when I use comfy? Or will comfy just fail or offload to CPU?

So, if having two cards pinned one to LocalAI and one to ComfyUI is a good choice, I can look into the hardware doubts. If nothing would practically change, no …

Understood, so my iGPU is too old, and i am better off with my NVIDIA only. Also, putting two might be just a waste, and i doubt my PSU can handle them even if i had the proper PCIe slot.

So, using llama.cpp if i use a model bigger than my VRAM will offload to CPU? I am not fully clear on this

It is as secure as it can be.

Are you planning to connect to the server from internet? Are you planning to expose services to the internet?

Next time use Gentoo. Your life will be ruined forever but I assure you none of those “defaults” are imposed… And you get plenty of error messages oh yeah

Not at all …

You are free to do as you please, and I fully respect that.

I was also a no AI coder, but somehow changed my mind slowly as I learned how to use PROPERLY the tool, which can be quite useful.

Learning how to use it has been fun too, so I suggest you give it a try if you haven’t done so yet.

The first risk is abusing it. The second risk is trusting it. And there are many more risks, but AI is a knife and not a pistol: there are good uses for it, but you must be careful and use it properly all the time.

Disclosure is needed, I agree.

Let’s say it feels complex, and the tags will not avoid the discussion in the comments anyway … but it’s a start so good for it

Woah …

This is overly complex.

As a dev that sometimes published something, and I don’t vibe code butnl who doesn’t use AI nowadays? That is way too much complex. And zero projects today don’t use AI in any forms blnot even to search or bugfix …

Can it load models straight from lhuggingface?

Never had mdadm causing that. I have a raid5 array that’s used only for backup once a day and it is mounted 24/7, and the disks are off for 23hba day more or less …

Inotify is the official kernel supported tool that uses the inotify API, again that is the official kernel API for monitoring files and folders.

Its not that immediate to use, specially because there are caveats but works reliably well and indeed get the job done. Why do you say fragile? There isn’t a more comprehensive approach.

Only issue, doesn’t work on NFS and Samba shares.

Go for a n100 or even one with an Atom CPU. Get as much ram as you can afford…